>Cheryl's tuning letter fall 2015 described a case where they
implemented MSU capping ... [snip]


Ooops, I just saw the some text vanished from my mail... sorry for the 
incpmpleted data..


I wanted to add that the CPU Activity Report shows the system (LPAR) being busy 
between ~30% and ~50% (physically), and the whole CEC is nowhere near running 
at the physical (12CP ) capacity limits. We're running with Group Capacity 
Limit, and are also nowhere near being capped.


So there should be enough spare capacity for the system (LPAR) to to use. But 
it does not. The 7 vertical low CPs are mostly parked or unparked for only a 
few percent. I wonder why MVS is not using more CPs and more Capacity for that 
job.




But then there is the interesting comment from Greg: RMF might be fooling me. I 
understood RMF to count a CP delay sample when it finds a ready work unit on 
the WUQ (dispatcher queue). My (probably incorrect) thinking is that a ready WU 
should get access to a CP in this situation (spare capacity / parked CPs). This 
is why I'm interested in the mechanism that influence WU queueing and selection 
when HiperDispatch is on, i.e. when there is one WUQ per processor node.


--
Peter Hunkeler



----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to